In the Kitchen with NanoMine

11/19/20 Podcast

If discovering and designing next-gen materials is like cooking, Duke engineer Cate Brinson is writing the materials cookbook.

podcast cover art: mixing bowl and whisk
In the Kitchen with NanoMine

Featuring

L. Catherine  Brinson Profile Photo
L. Catherine Brinson Profile Photo

L. Catherine Brinson

Donald M. Alstadt Chair of Mechanical Engineering & Materials Science, Sharon C. and Harold L. Yoh III Distinguished Professor

Transcript

Host:  Hello and welcome to Rate of Change a podcast from Duke Engineering dedicated to the ingenious ways that engineers are solving society’s toughest problems, I’m Elizabeth Witherspoon.

[Sound effects: Start of gas stove]

Top chefs know a pinch of this or a dash of that can make all the difference between a memorable meal or a bitter disappointment. It takes a lot of trial and error until just the right combinations come together. The same is also true when discovering and designing the materials we need to solve complex problems in medicine, electronics, transportation, and sustainable energy. In many ways, it is a lot like cooking.

[Sound effect: Music]

Cate Brinson:  There’s some really interesting things about polymer nanocomposites. These are polymers that have tiny amounts of nanoparticles in them, usually a very small loading, so it’s 99.5% polycarbonate and then 0.5% fairy dust.

[Sound effect: Fairy ping]

Host:  This is Cate Brinson, the Alstadt Chair, and a Yoh professor in the Thomas Lord Department of Mechanical Engineering and Materials Science at Duke and an expert in polymer nanocomposites. She is one of the many Duke materials scientists and engineers working at the forefront of design discovery and deployment of the materials needed to drive innovation.

Cate Brinson:  And the fairy dust can be anything. It can be silica. It can be titanium. It could be carbon nanotubes. It could be any flavor nanoparticle you want, right? And that fairy dust can also be chemically modified so it can have little surface groups on it that cause it to interact differently with each other and with the polymer that you’re putting it in.

Host:  Polymer nanocomposites are in important innovations already. They have replaced metals in many existing inventions because of their lighter weight or better performance, such as in modern dental fillings that can match the color of our teeth or parts that make automobiles lighter and more fuel efficient. They’re used in food packaging, biomedicine, and solar energy. And they will feature prominently in future innovations as a result of the basic materials science and engineering research Brinson and her colleagues are now doing.

Cate Brinson:  You can also make the sample many different ways. You can melt mix it. You can extrude it. You could hot press it. You could ball mill it. There’s all kinds of things you can do to make a sample. And all of those things impact the physical properties of this thing that you get out. And by physical properties I mean mechanical properties, electrical properties, dielectric properties, conductivity, toughness, strength, diffusion coefficient, so it’s any physical property of the material system, optical properties, even, could be anything.

[Sound effect: Wooden spoon stirring batter in a bowl. Sound via https://www.zapsplat.com]

Host: Back to our cooking analogy, think of the role certain ingredients play, not only in terms of adding flavor, but in the successful outcome of, say, baking a cake. Only a few teaspoons of baking powder added to the several cups of flour releases carbon dioxide gas bubbles in the wet batter, causing it to rise during baking and have that light, airy texture we like so much. Without it you wind up with a hard, dense, rubbery disappointment. However, besides the obvious differences between designing polymers and baking a cake is that we know what to expect when you add baking powder to cake batter, or if you leave it out. In materials research polymers compared to metals, for example, are highly unpredictable.

[Sound effect: Alert Bells]

Cate Brinson:  Polymers are messy, messy, messy things. They’re long-chain and tangled molecules. They’re very disordered. They’re moving all the time, and they interact with each other and with things that are mixed in like nanoparticles.

[Sound effect: Music]

Host:  But don’t think for one minute that means Brinson or any of her materials research colleagues are giving up on this long, tedious effort that typically takes 20 years from discovery of a new material until it hits the marketplace.

President Barak Obama:  We can do it faster. To help businesses discover, develop, and deploy new materials twice as fast. We’re launching what we call the Materials Genome Initiative.

Host:  In June, 2011, the Materials Genome Initiative, or MGI, began funding materials scientists and engineers through five federal agencies for steps all along the pathway of materials research from basic discovery through design and development to cut this timeframe in half along with the costs. The name references the Human Genome Project, the international scientific research project of identifying and mapping all the genes of the human genome from a physical and functional standpoint. Declared complete in 2003, use of this widely available resource has ushered in the era of precision medicine and continues to spur medical innovation.

[Sound effect: Music]

Brinson’s colleagues working with metals and other highly ordered substances referred to as “hard matter” have and continue to build such resources.

But how are they going to do this with polymers, which are so messy and unpredictable?

One way is by harnessing the power of artificial intelligence and machine learning.

Cate Brinson:  You want to enable discovery and design. And the traditional methods to understand materials is experimental, so doing these tedious experiments, slowly, carefully, one at a time. Or doing some computation, where there are some limitations and that are forward predictive. But there’s something profoundly different that we could do by applying machine learning and data science, if we have a lot of data, and we do have a lot of data. But, that data is not currently accessible.

For example, if I want to know something about a material, I’ll put some search terms into Google Scholar or Web of Science, and say, “I want to know the glass transition temperature of silica polystyrene nanocomposites.” Very specific.

But, what do I get out of this search? I get a list of 34,000-plus results. Many of which are relevant, some of which are irrelevant, but it’s just a list, and the links in that list lead to individual PDF files. And then usually, through many clicks later, on a single entry, you get to one PDF file and then you have to read that one PDF file and you have to go extract that glass transition temperature that you were seeking for that composite or for the 12 composites that were in that particular paper, and then write that one number down in your spreadsheet.

And I can’t do that for 34,000 items on the list. It’s crazy! I should be able to do a search like this and get all the data on a plot, right? Because if I could get data on a plot, I could also take that data and if I also know, in addition to each data point, I also know, oh, for example, how were they made? What were they composed of? What were the processing conditions? What was the room temperature? What are all the other variables.

If you know all of these things, then you can start putting machine learning models and data science to work and start discovering new things. And if you have all of this data fully accessible, then you can also create a design loop where you could start to design materials based on this wide swath of data that we have, but we don’t have accessible, that you just can’t get it in a form that you need right now.

So, it’s rich, and it’s all out there, but it’s really only accessible to us one tiny little piece at a time like breadcrumbs, right? And we want to put the whole cookie together, we don’t want these isolated little crumbs from all over.

So, we’ve built this platform, a data framework, NanoMine.

[Sound effect: Music]

Host: NanoMine is a platform specifically written for polymer nanocomposites to make this data readily available to other researchers. This is the goal of the Materials Genome Initiative – to build and make available the resources needed to accelerate the discovery, design and development of new materials. Think of it as the step before the chefs can author their cookbooks of recipes. It will be the reference guide for the recipe writers that explains what the different ingredients behave like in various combinations and under various conditions.

Cate Brinson:  We hire undergraduates in the summer to curate data. We’re also trying to promote self-curation and we’re working on semi-automated methods of curation, and that’s something I want to work more on, semi-automated curation.

All right, with that as a backdrop, what do we do with this platform? What do we put in it? Well, I’ve always been interested in polymer nanocomposites, and we and others have done lots of interesting work trying to understand the properties of these systems, both computationally with simulations, and also experimentally looking at what’s happening physically – creating data that we can curate into NanoMine.

One of the really important things in these systems is that everything is interacting. How they interact is still largely unknown. And so, part of what we explore in the experimental side of our work is to try to understand how the polymer chains behave differently near particles or away from particles. Or, what if I have two particles very close together, and that’s actually different that now the polymer behaves around just a single particle in isolation. And these experiments you have to do are very difficult, because you have to do them at the nanoscale. There’s lots of interesting details down there.

But that also means that when you go to try to understand what the properties of the bulk materials are, the bulk composites, it just doesn’t add up from the properties of the particle plus the properties of the polymer. It’s also this affected zone, this interfacial zone of which we still don’t know a whole lot. So, this additional dearth of information makes these systems even more challenging, but also super interesting for new applications of data science.

So, in the materials genome space, I call polymer nanocomposites my model materials system – analogous to the life sciences where you have your model organisms, like dictyostelium or drosophila or the mouse model. Those are your model organisms, and you focus in that space. So, polymer nanocomposites are my model organism, my model material. I call them the slime mold of materials. Because you think, “Slime mold, who cares?” But if you focus in the genome of that one organism and connect it to its properties, you can learn something fundamental with far-reaching impact. And so, the same thing is true here, in my slime mold of materials polymer nanocomposites. There are a lot of really interesting problems which will allow us to learn both fundamental mechanisms connected to control variables by accessing large swaths of data with AI. For example, what controls those interfacial properties? And, will allow us to develop intelligent design loops for unimagined materials with superb properties.

[Sound effect: Music]

Host:  Brinson is so passionate about the need for harnessing the power of artificial intelligence and machine learning to propel materials science forward, she is leading a three-year, $5-million National Science Foundation graduate training grant to train the next generation of both computer and materials scientists and engineers in these skills. The program is called aiM, shorthand for “AI for Understanding and Designing Materials.” It brings together PhD students in computer science with PhD students in materials science to cross-train in such a way that those steeped in AI skills learn about materials and those steeped in materials science learn how to use AI.

The future is all about team science in which multidisciplinary teams will work together to solve complex problems. This program helps bridge the gap as these two fields converge and it gives them common language and skills to work together more efficiently, and most of all, more successfully.

Thank you for listening today, we hope you’ve enjoyed it. If you have, please share it and subscribe to hear more stories about Duke engineers.

Keep Listening

Explore all seasons of the Rate of Change podcast, dedicated to the ingenious ways engineers are solving society’s toughest problems.